Re: DISTINCTness of literals

On Sep 25, 2006, at 11:22 AM, Seaborne, Andy wrote:

> We left open what to do about literals of the same value in  
> DISTINCT results.
>
> I propose that term-distinct apply to literals, and not a form of  
> value-distinct.
>
> SPARQL is defined for simple entailment anyway so for this  
> entailment regime it would be term-distinct for literals.

While this is a coherent view, it's not quite nailed down that  
comparison for the purpose of distinctness is tied to the entailment  
regime. Value testing, as I understand Eric, is not.

> I propose that rq24 only talk about this.
>
> Some considerations:
>
> 1/  It's terms that are returned, not values, when encoded into the  
> result set.  But which term should be returned if two literals are  
> value-distinct and not term-distinct?
>
> e.g.
>
> "1"^^xsd:integer
> "01"^^xsd:integer
> "1.0"^^xsd:decimal

I would give either the first in sort order or any (implementation  
defined), with a preference for the latter..

> The SPARQL test suite is based on RDF graph equality [1] which uses  
> term equality for literals.
>
> 2/ XPath/XQuery Functions&Operators [2] allows numeric type  
> promotion - but XML schema datatypes does not and double/float/ 
> decimal have different value spaces.  See [3].

One word, ew.

> "1.3"^^xsd:double != "1.3"^^decimal

This would indeed have to be decided. Of course, per usual, we could  
allow for parameterization.

Cheers,
Bijan.

Received on Monday, 25 September 2006 14:02:15 UTC